Discussion SpotOn - an insulin dosage prediction app

encore1332 said:

I will have a go, pancreas seems to have finally died of and stopped causing trouble. in the process of working out my ratios and any help would be appreciated

Click to expand...

Hi encore1332!

Thanks so much for your interest! However, I'm afraid that at the moment the app is in such an early stage of development that I would not recommend to use it when your dosage is not yet fine-tuned. Sorry, I hope you understand.

However, once the app is a bit more proven it will be able to help you fine-tune your dosage!

Hi everybody,

The app has a new update and I'm looking for more testers! Please contact me if you're interested!!

Learning is the future!

Hi there

Great! I'm happy to hear that you're interested!

You can download the app here. There is a manual, which is a bit outdated, but describes the necessary steps for installing the app here.

If you have any questions, just post them here, send me a message or send me an email via spotonapplication (at) gmail (dot) com. I'll be happy to help, get feedback and new suggestions! I'll try to get back to you within 24h.

As I said, this is currently under development, so I'm sure there is still plenty to do. Since I programmed this app myself, I probably won't see inconsistencies like you will, therefore I'd be really happy to receive some feedback from you!

Any thoughts on the app?

HI DunePlodder,

Great to hear that you like the project! Concerning your questions:

Yes, I think you're right. I will update the manual describing the set-up of the model! Also back button behaviour is something that I could work on. However, this is common to all android apps with the "drawer" layout.

DunePlodder said:

I'd like to see a prediction at say 15/30 minute intervals with a graph to make it more interesting.

Click to expand...

You actually can, by clicking on the "predicted value in 4 hours". Note that this value is based on the "learning algorithm". There are basically two prediction algorithms running in parallel in the app:
The first you set up by setting your model. It then uses the static formula you see when entering insulin to predict the required dosage.
The second uses all the values you entered and tries to learn the relationship between entered BG value, carbs and dose to estimate the factors. The estimated factors you can see by clicking on menu (top right) --> model output. It will only work, however, once you've entered a number of days (~5-10) and repeatedly updated the model

DunePlodder said:

Could you explain what the variables are in the Bolus Wizard formula?
Not sure how to use "Missed Entry".
The 4 hour results so far have been too low but I think I need to re think some of my parameters.

Click to expand...

1) two sets of variables each for every period of the day. Carbohydrate ratio: How many units of insulin do you need for each gram of carbs. Correction factor: How much does your BG value drop for unit of insulin.
2) Missed entry: please use it if you missed to enter something (carbs, insulin, sport) BEFORE entering the next BG measurement. This ensures that the prediction algorithm associates the cause-effect relationship properly.
3) This is normal at the beginning, but it should converge to a reasonable value within a couple of days. As said, this is based on the second model es described above!

Thanks so much for your feedback! I hope with that the answers above answer your questions. If not please ask!!!

And I'm really happy to hear further thoughts!

Yes, Value in 4 hours is from "now" and yes, using the "different time" was the correct thing to do! You only need to use "missed entry" if you have forgotten or are too lazy ;-) After 8 hours, the app is automatically going to assume that you forgot to enter something and adds the "missed entry" by itself. So addding values retrospectively should work just fine!

Update model: it updates the second (learning) model described in my post above. It's actually optimizing the detected cause-effect relationship. Please let me know if this takes too long, I've heard complaints from other users.

Reset model: it resets the second (learning) model to a default value. This funciton can be useful if the learning model predicts c**p ;-)

I still want questions by the way!

So what do you think? Are you still using the app? If yes, what would you like to have improved? If no, same question ;-)

Yes, true a different day cannot be selected. I could add the function without much difficulty, but I think that most users would not use it. If this is really important to you I would add it for testing purposes, as you suggested, and take it out later.

Just out of curiosity, what is your usual software?

Have a nice weekend!